What is Data Science? A Beginner's Guide with Real Examples

Imagine you run a sweet shop near Sarafa Bazaar. Every evening you note in a register how many plates of jalebi, poha and gajak you sold, and what the weather was like.

After three months, you have about 90 pages of numbers. One day you wonder: "Should I make more gajak on cold evenings?" You flip through the register, count cold days, compare sales, and decide to prepare 40 extra pieces whenever the temperature drops below 15 degrees.

You just did a tiny data science project.

What is data science, in plain words?

Let's map the sweet shop onto the real vocabulary.

• The register is your dataset. Each evening is a row, and each thing you wrote down (date, item, plates sold, temperature) is a column.
• "Should I make more gajak on cold evenings?" is the business question.
• Counting and comparing cold and warm days is analysis.
• "Make 40 extra pieces below 15 degrees" is the decision, the part that actually earns money.

Data science is the practice of using data, statistics and code to answer a real question well enough that someone can act on the answer.

The key word is act. A chart nobody uses, or a model that never reaches the shop counter, is not data science. The job ends when a decision changes.

Data science vs data analytics, machine learning and statistics

These four terms get mixed up all the time. The easiest way to separate them is by the question each one is best at.

Each neighbouring field answers one kind of question; data science picks whichever ones the business problem needs.

Here is how the sweet shop question looks through each lens:

• Statistics: gajak sold 52 plates on cold days and 31 on warm days on average. Is that gap big enough to trust, or could it be luck over 90 days?
• Analytics: a monthly report showing sales per item, per weekday, with a note that December was the best month.
• Machine learning: a model that takes tomorrow's forecast temperature and predicts plates sold, say 48.
• Data science: combining all three to set a preparation rule, checking it saves money, and revisiting it next winter.

So data science is not a separate island; it borrows from the other three. In small companies one person does all of it, while large companies split the work across several roles.

A tiny data science project you can run

Let's turn the sweet shop register into code. You don't need anything installed; the block runs in your browser. If Python itself is new to you, start with your first Python program and come back.

# Ten evenings from the shop register: temperature (deg C) and gajak plates sold
temps = [12, 14, 22, 25, 11, 13, 24, 27, 10, 23]
plates = [55, 50, 30, 28, 58, 49, 33, 26, 60, 31]

cold = [p for t, p in zip(temps, plates) if t < 15]
warm = [p for t, p in zip(temps, plates) if t >= 15]

print("Cold evenings:", len(cold), "average plates:", sum(cold) / len(cold))
print("Warm evenings:", len(warm), "average plates:", sum(warm) / len(warm))
print("Extra plates on cold days:", sum(cold) / len(cold) - sum(warm) / len(warm))

That is the analytics part: what happened. Cold evenings sold about 25 more plates. Now let's add a small predictive step with scikit-learn, the most common machine learning library in Python.

from sklearn.linear_model import LinearRegression

temps = [[12], [14], [22], [25], [11], [13], [24], [27], [10], [23]]
plates = [55, 50, 30, 28, 58, 49, 33, 26, 60, 31]

model = LinearRegression().fit(temps, plates)
print("Plates lost per degree warmer:", round(-model.coef_[0], 2))

for forecast in [9, 16, 26]:
    guess = model.predict([[forecast]])[0]
    print(f"Forecast {forecast} C -> prepare about {round(guess)} plates")

Every extra degree costs roughly two plates. The data science part is acting on that: prepare stock from the forecast, track waste for a month, and check the rule saved money.

Real examples of data science in India

Data science sounds abstract until you notice it in apps you already use.

• Food delivery: the estimated delivery time on your order is a prediction built from distance, restaurant prep history, traffic and rider availability.
• UPI fraud checks: banks score every transaction in milliseconds. An unusual amount, a new device and a late-night hour together can trigger an extra check.
• E-commerce recommendations: "customers who bought this also bought" is a pattern mined from millions of past baskets.
• Retail stock planning: a chain of stores in Indore deciding how much to stock before Diwali is solving the sweet shop problem at a bigger scale.

Every example starts with a decision (dispatch a rider, block a payment, order stock) and works backwards to the data.

Skills you need to become a data scientist

You don't need a computer science degree. You do need a mix of four skill groups.

Programming with Python

Python is the default language for data work. You need comfort with variables, loops, functions and lists, then the data libraries on top: NumPy for numbers and pandas for tables. Our guide to NumPy arrays is a good next step after the basics.

Statistics and maths

You need averages, spread, distributions, correlation and the idea of sampling. Most daily work uses school-level maths applied carefully, not advanced calculus. The statistics for data science tutorial covers the essentials with code.

Data handling and SQL

Real data lives in databases and spreadsheets, and it is messy. Expect to spend much of your time joining tables, fixing dates and filling gaps. Most jobs require SQL.

Business sense and communication

Beginners skip this one. You must turn "reduce wastage" into a measurable target, then explain the result to a manager who has no interest in your code.

# A quick self-check: how ready are you? Score 0-3 for each skill.
skills = {"Python": 2, "Statistics": 1, "SQL": 0, "Communication": 2}

total = sum(skills.values())
weakest = min(skills, key=skills.get)
print(f"Readiness: {total} out of {3 * len(skills)}")
print("Start with:", weakest)
for name, score in sorted(skills.items(), key=lambda kv: kv[1]):
    print(f"{name:<14}{'#' * score}{'.' * (3 - score)}")

If you would rather learn these skill groups in a classroom than alone, Debugshala's data science course is taught in classrooms at Old Palasia, Indore, and also online.

Common mistakes beginners make

• Starting with the model. People jump to deep learning before asking what decision the result will change. Start from the question.
• Trusting the data blindly. A temperature column with values like 250 is a typo, not a heatwave. Always look at minimums, maximums and missing values first.
• Confusing correlation with cause. Ice cream sales and drowning cases rise together because both rise in summer, not because one causes the other.

Key takeaways

• Data science turns data into decisions using statistics, programming and domain knowledge.
• Statistics checks whether patterns are real, analytics explains the past, machine learning predicts, and data science combines them around a business question.
• You can start with plain Python and lists, then move to pandas, statistics and simple models.
• The four skill groups are Python, statistics, data handling with SQL, and communication.
• Always begin with the decision and work backwards to the data.